DHCP don't acknowledges more than 80 users??
jw354 at cornell.edu
Fri Dec 5 17:29:19 UTC 2008
On Dec 5, 2008, at 2:49 AM, Martin Hochreiter wrote:
> Joe in MPLS schrieb:
>> IIRC... failover peers split the address pools for each subnet with
>> each peer getting half of the addresses. A peer won't claim the other
>> half of the addresses until its partner is down long enough to be
>> considered (more or less) permanently down. Do you have 160 addresses
>> in that subnet's pool?
> mhmm - that's a good hint -
> The dhcp serves about 10 subnet's, in the particular 10.222.xxx.xxx
> net are 272 clients.
> I turned down the secondary 12hours before to avoid that reclaiming
> problem - I thought
> the default value for reclaming is 30 minutes so that shouldn't be a
> problem if I kill the secondary
> 12hours before.
> Maybe I am wrong here
> dhcp-users mailing list
> dhcp-users at lists.isc.org
In communication interrupted state, the server will be serving roughly
half the unused addresses in the pool.
If you want that to be sufficient for your needs during an outage, you
can put enough addresses in the pool
so that half is sufficient for your needs (if you have that many
addresses). If you want a server to take over the other
server's work, you can put it in partner down state. If you want it to
take over automatically, you can script up whatever
tests you think are sufficient to assure you that the other server
truly is down (e.g., 30 minute wait, ping, info from a 3rd server that
monitors, etc) and have the script put dhcpd in partner down state.
Basically, dhcpd does nothing on its own that
could start allocating duplicate addresses, a potential disaster,
simply because of a lack of network connectivity
between the two dhcp servers. The other server could be running even
if the servers can't talk IP, e.g. a wrongly
installed acl in a router, and in such a case, the two servers are
designed to continue functioning without stepping on
each other's toes. The decision to assume the other server truly is
out of service is with you, or whatever scripted testing
you've decided sufficiently mitigates the risk for your needs.
If the server doesn't seem to be actually serving half the unused
addresses in the pool, you could inspect
the lease file to see what state the unused ips are in both before you
kill the other server and after.
More information about the dhcp-users